ZEILX.AI — Verification
Verification

How the sources
get checked

Every report on this site is only as good as the evidence under it. This section documents how that evidence is checked — the method, the reasoning behind it, and the tools being built to keep it honest over time.

J. Norman Dixon · Zeilx Independent Research
Method v1.1 · Living document · Revised July 2026

§ 01Why this section exists

A clean reference list looks identical whether every source in it was checked or none of them were. Nothing in a finished document reveals the difference. That is a problem for any reader trying to decide how much weight to give a claim — and it is a problem for the writer too, because a method held only in one's head drifts.

So the method is written down here, in public, in enough detail to be criticized. It is not an invention: it assembles established practice from intelligence analysis, evidence-based medicine, archival scholarship, and journalism into one sequence. The value, such as it is, lies in the arrangement and in the fact that it is disclosed.

Four passes

  1. Existence — is it real?
  2. Provenance — where did it originate?
  3. Corroboration — does anyone independent confirm it?
  4. Disconfirmation — what would prove it wrong?

Reliability tiers

T1Primary / authoritative T2Credible secondary T3Supporting / contextual T4Flagged / use with caution

§ 03Where it stands

The four passes are current working practice and are applied by hand to every report published on this site. The tiering and the appendix format are in use now.

The automation is not. The source ledger and its monitoring workflow are designed and under construction; nothing on this site should be read as claiming they are running yet. When they are, this page changes and the build gets documented.

An open weakness, stated rather than buried. The T1–T4 scale grades a source on one axis. Intelligence practice — the Admiralty Code — grades source reliability and information credibility separately, on the reasoning that a reliable source can still report something wrong. This method currently compensates by tiering per claim rather than per outlet. Whether to adopt a second axis is unresolved, and is the most likely change in the next revision.